←
Nutrient5,341 posts
Nutrient avatar

Nutrient

@nutrientdocs

The deterministic document platform for agents and humans. SDKs, cloud APIs, integrations, and workflow automation turn documents into reliable workflows.

◷ Joined August 2012▧ 1,038 Photos and videos
Nutrient avatar
Nutrient@nutrientdocs·6h

The details matter in document parsing. This WholeDoc Bench case takes a closer look at edge cases which make most parsers fail. We put our parser head to head with frontier models, and the results were clear: Nutrient ourperformed frontier models by a significant margin. Review the benchmark: bit.ly/4kZ4KWE

2:39
Nutrient avatar
Nutrient@nutrientdocs·Nov 16

What is a decision model? Learn in under 60 seconds. We sat down with AI Engineer Austin Nguyen to discuss what decision models are, and how they're used. Learn more about the Nutrient decision model: bit.ly/4jpIZyF

0:41
Nutrient avatar
Nutrient@nutrientdocs·Nov 13

What happens when a table runs across multiple pages? Most parsers will break. Multi page documents are not something you can reliably extract with most tools, and that's a problem. But not for us. We tested a bank statement with 222 transactions, and reliably extracted the correct information every time. See the example: bit.ly/4C26nsW

0:18
Nutrient avatar
Nutrient@nutrientdocs·Nov 13

How can you search through thousands of documents? Try to give it to an agent and you'll overwhelm the context. We discuss NuDocs, an AI-native document editing suite build specifically to solve problems like this.

1:38
Jonathan Rhyne avatar
⟳ Nutrient retweeted
Jonathan Rhyne@jdrhyne·Nov 12

We built a parsing benchmark WholeDoc Bench around where documents actually break: tables that run across pages, paragraphs split by a page break, scans with typos that have to stay typos. Nutrient is #1 overall. The best frontier model scores 67.1 at 10x our price per page. Only 2 of 19 parser settings get both page-break cases right. Both are ours. Code is Apache 2.0. Think your parser belongs higher? Open a PR.

Today we're releasing WholeDoc Bench, a benchmark that scores document parsers on whole documents, not just individual pages. Nutrient topped the leaderboard at just a tenth of the price compared to frontier models: • Nutrient Agentic: 86.1/100 at 1.8¢ per page • Best frontier model: 67.1/100 at about 18¢ per page Here’s how it works 🧵

Post attachment
Nutrient avatar
Nutrient@nutrientdocs·Nov 12

We built an AI sales agent, and now we're sharing the playbook. 🚀 Our CRO Richard Malloy and Director of PMM Greg Ives joined SalesforceBen to walk through the build and share 4 rules for bringing AI into your own org. Watch on demand:bit.ly/sfben-webinar

Nutrient avatar
Nutrient@nutrientdocs·Nov 12

If your AI agents get even 10% of its data extraction wrong, can you really trust it at all? We discuss why reliability is so important when working with your documents.

1:19
Nutrient avatar
Nutrient@nutrientdocs·Nov 10

What we found: tables that continue onto the next page break most parsers. Fewer than 1 in 5 kept them in one piece. Even higher reasoning levels were not an effective solution. We ran tests with multiple reasoning levels, which moved scores less than half a point.

Nutrient avatar
Nutrient@nutrientdocs·Nov 10

Why whole documents? Many parser errors happen between pages. Most parsing benchmarks grade one page at a time. So a table that runs across 5 pages gets graded as separate documents, and nothing checks that it comes back as one table. When a sentence carries over to the next page, many parsers paste the page number and header into the middle of it: “…could fund a mobile Page 3 · Annual Report 2025 repair van.” Graded page by page, both halves look fine. WholeDoc Bench grades 13 parsers on the whole document, with the same rules for all of them.

Post attachment
Nutrient avatar
Nutrient@nutrientdocs·Nov 10

Today we're releasing WholeDoc Bench, a benchmark that scores document parsers on whole documents, not just individual pages. Nutrient topped the leaderboard at just a tenth of the price compared to frontier models: • Nutrient Agentic: 86.1/100 at 1.8¢ per page • Best frontier model: 67.1/100 at about 18¢ per page Here’s how it works 🧵

Post attachment
Nutrient avatar
Nutrient@nutrientdocs·Nov 5

Most AI tools handle 90% of documents fine. The last 10% is scanned pages, messy formats and tracked changes, and that’s where you stop trusting the answer. We sat down with Nick Winder to talk about NuDocs, the AI-native tool built for that last 10%. Try it on macOS: nudocs.ai

12:35
Nutrient avatar
Nutrient@nutrientdocs·Oct 23

Fast answers, straight from the document. Watch the mailroom cat make a decision with Nutrient Data Extraction.

0:22
Nutrient avatar
Nutrient@nutrientdocs·Oct 22

Every mailroom needs a good hire. This one sorts documents before you finish your coffee. 🐾📑 Nutrient’s classify tooling sorts incoming documents by type so the right document reaches the right team, without anyone sorting by hand.

0:26
Nutrient avatar
Nutrient@nutrientdocs·Oct 20

What is the Nutrient Decision Model, and why would you ever use it over a large language model? Easy answer: cost and speed

0:43